arXiv:1604.

07422v1 [quant-ph] 25 Apr 2016

Single-world interpretations of quantum theory

cannot be self-consistent
Renato Renner

Daniela Frauchiger

Institute for Theoretical Physics

ETH Zurich, Switzerland

Abstract
According to quantum theory, a measurement may have multiple possible outcomes. Single-world interpretations assert that, nevertheless, only one of them
really occurs. Here we propose a gedankenexperiment where quantum theory
is applied to model an experimenter who herself uses quantum theory. We find
that, in such a scenario, no single-world interpretation can be logically consistent.
This conclusion extends to deterministic hidden-variable theories, such as Bohmian
mechanics, for they impose a single-world interpretation.

Introduction

Imagine an experimenter who reads notes about an experiment that she carried out
in the past. According to the notes, she measured the vertical spin component of an
electron that was initially polarised along the horizontal direction
p
p
|i = 1/2 |i + 1/2 |i .
The measurement had two possible outcomes, z = 21 and z = + 12 , corresponding to
projectors along |i and |i, respectively. Unfortunately, our experimenter long forgot the
details of the experiment, and the page of her lab notebook containing the measurement
outcome got lost. Still, she may apply quantum theory and find that the probabilities
associated to the two possible outcomes are
P [z = 12 ] = P [z = + 12 ] = 1/2 .

(1)

But this statement is manifestly symmetric it does not give any preference to either
z = 12 or z = + 12 .
Suppose now that our experimenter gets told that the measurement outcome z = 21
must have occurred. Given this piece of information, she may be tempted to conclude
that also the following statement is true.
Outcome z = 21 occurred whereas outcome z = + 12 did not.
renner@ethz.ch

(S1)

z = -

|
0:00

0:01

Figure 1: A basic quantum measurement experiment. A source emits an electron polarised according to . The detector measures the vertical polarisation direction z.
The starting point of this work is the (well known) fact that this conclusion is not the
only logically possible one. Noting that the statement outcome z = + 12 occurred is
not the logical negation of outcome z = 21 occurred, the following assertion could as
well be an accurate description of reality.
Both outcome z = 12 and outcome z = + 21 occurred.

(S2)

Hence, to conclude that (S1) is the correct statement, an additional assumption is necessary. The assumption could be that there is only one single world, in the sense
that among the different physically possible outcomes of a measurement, only one is
real.1 It contrasts with a many-worlds assumption, according to which all possible
outcomes are equally real even though we are always only aware of one of them. The
latter would imply (S2), thus retaining the symmetry of the probabilistic expression (1)
obtained from quantum theory.
Physical theories that impose that quantum measurements have single outcomes are
often called single-world interpretations. (The term interpretation alludes to the fact
that textbook quantum mechanics leaves open the question of how an expression such
as (1) should be translated to a statement such as (S1) or (S2). It is hence necessary
to interpret (1).) The main contribution of this work is an argument which, adopting
this terminology, asserts that single-world interpretations of standard quantum theory
cannot be self-consistent. More specifically, we prove a theorem that corresponds to the
following informal claim.
Main result (informal version)
There cannot exist a physical theory T that has all of the following properties:
(QT) Compliance with quantum theory: T forbids all measurement results that are forbidden by standard quantum theory (and this condition holds even if the measured
system is large enough to contain itself an experimenter).
(SW) Single-world: T rules out the occurrence of more than one single outcome if an
experimenter measures a system once.
(SC) Self-consistency: T s statements about measurement outcomes are logically consistent (even if they are obtained by considering the perspectives of different experimenters).
1 For our argument it will not be necessary to define the term real. Relevant is merely that, in a
single-world view, one particular measurement outcome must be singled out.

Property (QT) refers to standard quantum theory, which does not impose any constraints on the complexity of objects it is applied to [Deu85]. It is certainly legitimate to
question whether a theory that accurately describes nature must respect this requirement
after all, we are still lacking experimental evidence for the validity of quantum theory
on macroscopic scales. Conversely, self-consistency, as defined by (SC), is arguably an
unavoidable requirement to any reasonable physical theory. We are thus left with two
scenarios, depending on the outcome of future experiments.
Scenario 1: Experiments reveal that quantum theory is inaccurate in certain
regimes and must be replaced by a different theory. (It could be, for instance,
that we discover a yet unknown physical mechanism that leads to an objective
collapse of the wave function.)
Scenario 2: Experiments confirm that quantum theory accurately describes systems as complex as experimenters who themselves apply quantum theory. In this
case, the result proved here forces us to reject a single-world description of physical
reality.
A nice example that illustrates the use of the main result presented here is the de
Broglie-Bohm theory, also known as Bohmian mechanics or pilot-wave theory [DB27,
Boh52, DT09]. According to this theory, measurement outcomes are determined by
particle positions, which are well-defined at any time, so that (SW) is satisfied. Bohmian
mechanics is also compatible with quantum theory (without a collapse mechanism), so
that (QT) holds, too. We must therefore conclude that it cannot satisfy (SC). That is, if
Bohmian mechanics is applied from different experimenters perspectives, one sometimes
obtains mutually contradicting statements. In fact, the same conclusion holds more
generally for any deterministic hidden variable theory.
The paper is organised as follows. In Section 2, we introduce a (rather minimalistic)
framework to formally capture the relevant aspects of different interpretations of quantum theory, such as single-world vs. many worlds. The main result is stated in Section 3.
For its proof, we introduce in Section 4 an extension of the Wigners Friend gedankenexperiment [Wig67], which is then analysed in Section 5. In Section 6 we discuss specific
theories that have been proposed, such as Bohmian mechanics, in the light of the results
presented here.

2
2.1

Framework: Physics in terms of stories

Physical theories as constraints on stories

The main result, as described informally in the introduction, refers to the notion of a
physical theory T , and we should therefore spell out what we mean by this. Yet, all we
need is a characterisation in terms of certain necessary criteria. The reason is that the
result asserts the non-existence of a theory T with certain properties. Hence, the looser
the definition we use for characterising T , the stronger is the claim (cf. Section 6.4).
The approach we take is based on the paradigm that any physical theory T comes with
a set of laws that forbid certain things from happening. For example, thermodynamics
forbids that heat flows from a cold to a hot body. Another example would be a collapse
theory, which forbids that a macroscopic object can be in a superposition of being at two
3

forbidden
by ED

forbidden
by QT

Figure 2: Physical theories. is the set of all stories. Physical theories, such as quantum
theory, QT, or electrodynamics, ED, rule out some of them.
distant locations. This paradigm connects to the idea of falsifiability. If an experiment
leads to an observation that is forbidden by T then we have falsified T . Conversely, if
nothing is forbidden by T then T cannot be experimentally falsified.
A basic ingredient to our approach is the notion of a story, i.e., an account of events
that occur. Specifically, we consider stories, real or fictitious, about experiments such
as the spin measurement described in the introduction (see Fig. 1). To expand on
this example, suppose that the electron is emitted by a source that polarises it along a
direction as indicated by the position of a knob. Right after, its vertical spin component
is measured and the outcome z is shown on the display of the measurement device. In
addition, there shall be a clock that shows time t. A possible story may then read as
follows.
n
At t = 0:00 the source is invoked with knob position = |i and at
s1 =
1
t = 0:01 the measurement device shows outcome z = 2 .

Another story could be

n
At t = 0:00 the source is invoked with knob position = |i and at
s2 =
1
t = 0:01 the measurement device shows outcome z = 2 .

Many more stories are conceivable of course, and we will in the following denote by
the set of all possible stories. For our purposes, it is not necessary to characterise this
set further it suffices to simply postulate its existence.2 Yet, for concreteness, one
may think of as the set of all finite sequences of letters or of all sequences of English
words. Most of the stories then have no well-defined meaning, but this is unproblematic
as we shall see later.
Given a story s , we may find that it contradicts certain laws of physics. In this
sense, physical theories impose constraints on the set of all possible stories (Fig. 2).
For example, we would say that quantum mechanics forbids story s2 , but not s1 . This
motivates the following characterisation of physical theories.
A physical theory, T , specifies a subset of , whose elements are called forbidden stories.
We write s
/ T to indicate that s is forbidden by T .
The terminology is deliberately chosen in an asymmetric manner. If a story s is
not forbidden by a physical theory T , then this should not be interpreted as s is allowed
2 In

certain contexts it is useful to impose that the set be countable, though.

by T . For example, it could be that s is not precise enough for T to be applicable, or it

could be that s simply has no meaning. (Recall that we did not impose any constraints
on the set of all stories .)

2.2

Interpreting stories in terms of their plots

A story, in the usual sense of the word, has a plot, corresponding to the series of events
that happen according to it. We will use this notion to specify, in a precise manner, what
a story s tells us about an experiment. To illustrate this, we may once again consider
the spin measurement experiment of the introduction. This experiment has the following
generic structure.
Basic Measurement Experiment (BME)
Given a quantum system with Hilbert space H and a family of measurement projectors {z }zZ on H, perform the following steps at the corresponding times t.3
@ 0:00
@ 0:01
@ 0:02

Prepare the system in state .

Measure the system w.r.t. {z }zZ and record the outcome z.
Halt.

When carrying out such an experiment, which we may call E, we are typically interested in a well-defined set of events that may potentially occur. We denote this set
by [E]. In the case of the spin measurement experiment, any event of interest could,
for instance, be identified with a particular reading of values shown by the clock, the
source (whose knob position shows the prepared state), and the measurement device.
This would motivate an event space of the form


[E] = (t, , z) T H (Z ) ,
(2)
where indicates that no outcome has been recorded, and where


T = n:m : n N0 , m {0, . . . , 59}

(3)

is the set of possible time values t shown by the clock. This latter set is discrete, capturing
the idea that the clock has a finite precision.
Given the set [E] of potential events that can happen in the experiment E, a story s
about E now defines a plot sE , i.e., a subset of events that actually occur according to
the story. For story s1 from Section 2.1, for instance, the plot could be taken to be


sE1 = (0:00, |i, ), (0:01, 0, 21 ) .
(The assignment = 0 shall indicate that the measurement is interpreted as a destructive
one.) Similarly, for s2 , a natural choice could be


sE2 = (0:00, |i, ), (0:01, 0, 12 ) .
3 Here and in the following, times indicated in an experiment should be understood as placeholders
that can be substituted by any other points in time with the same order. Furthermore, we assume that
the systems states remain unchanged unless the protocol prescribes a specific action on them. In the
present example, the evolution of the systems state between t = 0:00 and t = 0:01 shall hence be trivial.

The plots sE1 and sE2 thus correspond to interpretations of the stories s1 and s2 , which
were described in prose, declaring precisely which events happen according to them.
This idea is captured more generally by the following definition.
An experiment, E, specifies a set of events, denoted by [E], as well as a function that
assigns to certain stories s a subset sE of [E], called plot of s.
We note that, formally, there is no constraint on how a story s is being interpreted
in terms of a plot sE in principle sE can be chosen arbitrarily. For our purposes,
it will only be relevant that the interpretations of a story in different experiments are
compatible with each other, in a sense that we are going to describe in Section 2.3 below.
In general, a story s could be ambiguous about whether or not a certain event in
[E] occurs. It could also be that s does not talk about the experiment E at all. In such
cases one may of course not want to assign a plot sE to s. The above definition accounts
for this, as it does not demand that sE exists for all s . In other words, s 7 sE is
generally a partial function on .
Importantly, the notion of a plot allows us to formally capture different interpretations
of quantum theory. While, for instance, the stories s1 and s2 above are in agreement
with a single-world interpretation, one could also consider other stories, according to
which the measurement provokes a branching, with different outcomes being observed in
the different branches. For example, we may define a story s1 with plot


sE1 = (0:00, |i, ), (0:01, 0, 21 ), (0:01, 0, + 21 ) ,
i.e., both outcomes z = 12 and z = + 12 occur according to this story.

2.3

Different views on the same story

Two experimenters may have different views on the same experimental setup and therefore describe it in two different ways. Consequently, their interpretation of a given story
s in terms of its plot will in general also be different. Nevertheless, the two plots must
of course still be compatible.
To explain what we mean by this, we first consider a simple situation where an
experiment E, as viewed by one experimenter, is a part of a larger experiment E0 , as
viewed by another experimenter. An example of such a larger experiment could be
an extension of the spin measurement experiment from above where, in addition to the
measurement at time t = 0:01, a second spin measurement with outcome z 0 is carried out
at t = 0:02. The event space [E0 ] of E0 may therefore be defined as the set of quadruples
(t, , z, z 0 ). However, when describing E we may ignore the second measurement, so that
[E] consists of triples (t, , z).
0
Formally, the two experiments are defined by (partial) functions s 7 sE and s 7 sE
from the set of stories to [E] and [E0 ], respectively. The functions specify how a given
0
story s should be interpreted. While the plots sE and sE may in principle be defined
arbitrarily, we want to model that E is a part of E0 . This is done by imposing certain
compatibility constraints. For instance, we may demand that any event (t, , z, z 0 ) that
occurs in experiment E0 must correspond to an event (t, , z) that occurs in the part E.

This could be captured by the compatibility constraint

(t, , z) sE

z 0 : (t, , z, z 0 ) sE ,

which should hold for any story s for which sE is defined. To simplify the notation, we
will in the following abbreviate terms of the form y : (x, y) by (x, ). The compatibility
constraint can then be rewritten as
(t, , z) sE

(t, , z, ) sE .

(4)

More generally, two experiments, E1 and E2 , may just have an overlap, but not be
contained in each other. Suppose for example that both of them include a measurement
of time t as well as some quantity z, but in addition also separate quantities, z1 and z2 , so
that their event spaces, [E1 ] and [E2 ], consist of triples (t, z, z1 ) and (t, z, z2 ), respectively.
The compatibility constraint that models that the two experiments have the quantities
t and z in common could then be formulated as the condition that
(t, z, ) sE1

(t, z, ) sE2

(5)

holds for any story s for which both sE1 and sE2 are defined.
We stress that, in the logic of the framework, compatibility constraints such as (4)
and (5) are not derived relations, but rather define how the different experiments are
related. In other words, (4) defines what we mean when we say that experiment E is a
part of experiment E0 . Similarly, (5) defines what we mean when we say that E1 and E2
have the quantities t and z in common.

2.4

Representation of physical laws

Recall that physical laws correspond to rules that forbid certain stories. To formulate
these rules, it is often convenient to consider specific experiments. The rules then take the
form of conditions on the plots of the stories, which can be expressed in mathematically
precise terms.
To illustrate this, consider the quantum-mechanical law that the measurement of a
quantum system with respect to projectors {z }zZ will with certainty give outcome z if
the prepared state satisfies kzk = 1. To formalise this law, let BMEH,{z } be the set
of all experiments E that have the form of the Basic Measurement Experiment introduced
in Section 2.2. Formally, their event space [E] is defined by (2), and the introductory
spin measurement example would correspond to the case where H = span{|i, |i} and
where 12 and 21 are projectors along |i and |i, respectively. The law that z occurs
if kzk = 1 can then be written as
(0:00, , ) sE and kzk = 1

(0:01, , z) sE .

(6)

That a theory T contains this law now means that the implication
(6) is violated for some E BMEH,{z }

s
/T

(7)

holds for any story s. Applying this criterion to the above examples, we find that s2 is
forbidden, whereas s1 and s1 are not.
7

The rule (7) manifestly depends on the set of experiments BMEH,{z } . This set
therefore defines the range of applicability of the corresponding physical law. It can
in principle be chosen arbitrarily. For example, if one holds the position that quantum
mechanics is only valid on microscopic scales, one may restrict BMEH,{z } to experiments where the measured system is small in some sense. Conversely, if one assumes
that quantum theory extends to macroscopic systems then the set should also include,
for instance, arbitrary quantum measurements on cats [Sch35].
In Section 3 we will use a generalisation of (7) to express formally what we mean when
we say that a theory T complies with quantum theory. The generalisation will include
experiments where the systems state evolves non-trivially and where the measurement
may be repeated.

Main result

The main result, which will be stated as Theorem 1 below, refers to three properties that
a physical theory may, or may not, have. In the following we describe them using the
story-based approach introduced in Section 2.

3.1

Compliance with quantum theory (QT)

Property (QT) corresponds to the assumption that the laws of standard quantum theory
are valid. In fact, for our purposes, it suffices to restrict to some particular rules that
are implied by standard quantum theory, i.e., we do not need to provide a full characterisation of the theory. To specify these rules, we consider the following experiment.
Repeated Measurement Experiment (RME)
Given a quantum system with Hilbert space H, a unitary U on H, a family of
measurement projectors {z }zZ on H, and a value z Z, repeat the following
steps for increasing n N0 until the halting criterion is satisfied.
@ n:00
@ n:01
@ n:02

Prepare a quantum system in state .

Let the system evolve according to U .
Measure the system w.r.t. {z }zZ and record the outcome z.
Halt if z = z.4

To simplify the notation, we will in the following represent the measurement projectors
{zH }zZ in the Heisenberg picture, i.e.,
zH = U z U .

(8)

While standard quantum theory generally makes probabilistic predictions, it also

implies certain deterministic statements. In particular, the following must hold for any
fixed z Z.
4 The halting step is not strictly needed. But since we are mostly interested in the case where 6= 0,
z

and since (QT) then implies that the experiment will end after finite time, it avoids the need for stories
that talk about an infinite sequence of measurements.

(a) If the prepared state satisfies kzH

k = 0 in round n then the outcome z = z

does not occur in this round.
(b) If the prepared state satisfies kzH

k = 1 in round n then the outcome z = z

occurs in this round.
(c) If the prepared state satisfies kzH
k 6= 0 and is identical in all rounds n until
the halting criterion is satisfied then the outcome z = z occurs at some point.
The definition of (QT) below captures the idea that any theory T that complies with
quantum theory must satisfy these rules.
To formalise the rules within our framework, let RMEH,{zH } be the set of all experiments of the type above, for any given Hilbert space H and a family of measurement
projectors {zH }zZ in the Heisenberg picture. On the formal level, we demand that the
event space of any E RMEH,{zH } is of the form (see also (2))


[E] = (t, , z) T H (Z ) ,
where T is defined by (3). The set RMEH,{zH } may otherwise be chosen arbitrarily,
provided that it includes the relevant parts of the Extended Wigners Friend Experiment defined in Section 4 below. This requirement will be captured formally by the
conditions (12), (14), (16), and (18).
The properties of a story s that are relevant for the formulation of (QT) can be written
as conditions on the plots sE for E RMEH,{zH } . For example, that the experiment is
repeated until the halting condition is satisfied corresponds to the implication
n = 0 or ((n 1):01, , z) sE for z 6= z
=

t [n:00, n:01] : (t, , ) sE . (9)

For convenience, we also define the set

n (s) = { : (n:00, , ) sE } ,
telling us which states are prepared in round n according to story s. (If s talks about
many worlds, for instance, this set could contain more than one element.)
T satisfies (QT) means that T forbids all stories s according to which in some experiment
E RMEH,{zH }z and for some z Z one of the following happens.
(a) In some round n, any n (s) satisfies kzH
) sE holds.
k = 0. Yet (n:01, , z
(b) In some round n, some n (s) satisfies kzH
) sE does
k = 1. Yet (n:01, , z
not hold.
(c) The experiment is repeated, i.e., (9) is satisfied, with n (s) = {} for fixed such
that kzH
6 0. Yet (n:01, , z) sE holds in no round n.
k =
Note that, according to standard quantum theory, all these conditions hold with
certainty.
9

3.2

Single-world (SW)

Property (SW) captures the idea that a measurement on a quantum system has only
one single outcome. More specifically, we demand that this is the case for all quantum
measurements relevant to one particular experimenter. (In the language of [Bru15], this
could be regarded as the requirement that there are facts maybe not per se, but
at least relative to an observer.)
To formalise this property, let Obs be a set of experiments of the form of the Repeated
Measurement Experiment as defined in Section 2.2. That is, any E Obs is also an
element of RMEH,{zH } , where H is a Hilbert space and {zH }zZ a family of projectors.
Crucially, the set Obs may be restricted to those experiments that describe the viewpoint
of one single experimenter. In the Extended Wigners Friend Experiment, this will be
Wigner, i.e., we will demand that the experiment W, defined in Section 4.2, is included
in this set (cf. (18) below).
T satisfies (SW) means that T forbids all stories s according to which for some experiment
E Obs the set


z Z : (0:01, , z) sE
has more than one single element.
We remark that (SW) is fundamentally different from the requirement that a theory T
be deterministic. For example, T may prescribe that the outcome of a spin measurement
is either z = 21 or z = + 12 (not both), but still assert that this outcome is not correlated
to anything that can be known before the measurement is carried out.

3.3

Self-consistency (SC)

Property (SC) is the requirement of self-consistency. Generally speaking, it demands

that the laws of a theory T do not contradict each other. To express this formally, let
Exp be the set of all experiments that the theory should be able to talk about. For us, it
is sufficient to demand that Exp includes the experiments F1, F2, A, and W, as defined
in Section 4 below.
T satisfies (SC) means that the condition
sE is defined for all E Exp
does not imply that s is forbidden by T .
To illustrate this, let E1 and E2 be the two experiments described in Section 2.3.
Suppose that a theory T forbids any story s according to which z1 = 12 in experiment E1 ,
as well as any story according to which z2 = + 12 in experiment E2 . If, in addition, T
requires that z = z1 holds in E1 and that z = z2 holds in E2 then there is obviously no
story s left that satisfies all rules of the theory. Hence, assuming that E, E0 Exp, the
10

theory T violates the self-consistency condition (SC), i.e., it rules out any possible story
about the two experiments.

3.4

Theorem

We are now ready to state the main claim.

Theorem 1. No physical theory T can satisfy (QT), (SW), and (SC).

Extended Wigners Friend Experiment

The proof of Theorem 1 is based on an extension of the Wigners Friend gedankenexperiment [Wig67], which is similar to the Schrodingers cat experiment [Sch35]. Wigner considered an experimenter, the friend, who carries out a measurement while being enclosed
in a perfectly isolated lab. The time evolution of the experimenters state (including the
lab) then corresponds to that of a closed system. The extension is inspired by Deutschs
idea of undoing this measurement and verify whether the original state can be retrieved [Deu85]. The specific setup considered here is similar to one proposed recently
by Brukner, which involves multiple experimenters who carry out a Bell test [Bru15].5
As a core ingredient we use a construction that originates in work by Hardy, where it
was used to establish Bells theorem without inequalities [Har92, Har93].6
We note that the purpose of the experiment is to prove Theorem 1. We therefore
do not have to worry about its technological feasibility at this point. We only need
to ensure that none of the steps of the experiment are forbidden by the basic laws of
physics. The experiment has hence a similar status as, for instance, the Schrodingers
cat gedankenexperiment. Nevertheless, a reader being worried about experimenters (or
cats) enclosed in perfectly isolated labs may substitute them by quantum computers
which simulate their actions.
In Section 4.1 we first describe the Extended Wigners Friend Experiment informally.
We will do this in terms of an experimental protocol that prescribes the actions of the
different participating experimenters from an overall perspective. Then, in Section 4.2,
we provide a formal characterisation using the story-based framework introduced in
Section 2. For this, we specify the views of the different experimenters individually.

4.1

A birds eye view on the experiment

The Extended Wigners Friend Experiment features four experimenters (see Fig. 3). Two
of them, the friends F1 and F2, shall be located in separate labs. We assume that, from
an outside perspective, they can be treated as isolated quantum systems unless there
are explicit communication steps in the experimental protocol. This means that the time
5 As

discussed at the end of Section 6.2, the argument presented here leads to different conclusions,
though.
6 Specifically, the global state (30) between F1 and F2 corresponds to Eq. 1 of [Har93] for the case

where 2 = (3 + 5)/6.

11

evolution of their state is described by a unitary. We also assume that F1 is fully informed
about the state of F2 (including F2s entire lab) at the time when the experiment starts,
and that this state is therefore pure. The other two experimenters, Wigner W and his
assistant A, are not only informed about the initial states of F1 and F2, but can also
carry out arbitrary quantum measurements on them.
Extended Wigners Friend Experiment (EWFE)
Repeat the following steps for increasing n N0 until the halting criterion in the
last step is satisfied.
@ n:00

F1 invokes a quantum random number generator and memorises the

output r {head, tail}.

@ n:10

Depending on whether r = head or r = tail, F1 sets the spin S of an

electron to state |iS or |iS , respectively, and hands it over to F2.

@ n:20

F2 measures S with respect to the basis {|iS , |iS } and memorises

the outcome z { 12 , + 12 }.

@ n:30

A measures F1 with respect to a basis {|okiF1 , |failiF1 } and records

the outcome x {ok, fail}.

@ n:40

W measures F2 with respect to a basis {|okiF2 , |failiF2 } and records

the outcome w {ok, fail}.

@ n:50

A and W compare their outcomes and halt if x = w = ok.

The quantum random number generator used in the first step shall be designed such
that the probabilities of the outcomes r = head and r = tail are 1/3 and 2/3, respectively.
For concreteness, we may think of a mechanism that prepares a qubit C, the quantum
coin, in state
p
p
(10)
C0 = 1/3|headiC + 2/3|tailiC
and measures it with respect to the orthonormal basis {|headiC , |tailiC }. Furthermore,
for the spin state of the electron prepared in the second step, which lives in a space HS
spanned by the two orthonormal vectors |iS and |iS , we use the convention
p
p
|iS = 1/2|iS + 1/2|iS .
The basis {|okiF1 , |failiF1 }, with respect to which the measurement by A on F1 is carried
out, shall be defined as
p
p
|okiF1 = 1/2|headiF1 1/2|tailiF1
p
p
|failiF1 = 1/2|headiF1 + 1/2|tailiF1 ,
where |headiF1 and |tailiF1 are the states of F1 (including her lab) at time t = n:30
depending on whether she has seen r = head or r = tail, respectively. Similarly, the basis
{|okiF2 , |failiF2 } of the measurement carried out by W on F2 shall be defined as
p
p
|okiF2 = 1/2| 12 iF2 1/2|+ 12 iF2
p
p
|failiF2 = 1/2| 12 iF2 + 1/2|+ 12 iF2 ,
12

z = +

w ok

r=tail
F1

x = ok
A

halt!
r = tail

z = +

F2

n:00

n:10

n:20

n:30

w = ok
W

n:40

n:50

Figure 3: Illustration of the Extended Wigners Friend Experiment. In any round n of

the experiment, one of Wigners friends, F1, polarises an electron depending on a random
value r. The other friend, F2, measures its vertical polarisation z. Wigners assistant,
A, and Wigner, W, measure the entire labs of F1 and F2, giving outcomes x and w,
respectively. The experimenters use quantum theory to derive statements about values
seen by others. F1, for instance, tries to infer w as measured by W. The experiment
ends when x = w = ok.
where | 12 iF2 and |+ 12 iF2 are the states of F2 at time t = n:40 in the case where z = 21
and z = + 21 , respectively.
Following the spirit of the approach described in Section 2, we will consider stories
that one can tell about this experiment. An example could be a story as follows.

At time t = 0:00 the output of F1s random number generator is

r = tail. She therefore prepares the electron spin S in state |i. When
F2 measures S at t = 0:20, she gets outcome z = + 21 . In their subses=

quent
measurements at times t = 0:30 and t = 0:40, A and W obtain

outcomes x = ok and w = ok, respectively. They therefore halt.

4.2

The experimenters views

To describe the Extended Wigners Friend Experiment formally, we subdivide it into

four sub-experiments, which we label by the four experimenters, F1, F2, A, and W.
The idea is that experiment F1 is a specification of the part relevant to experimenter
F1, and so on.7 The idea behind this subdivision is that each of the four experiments
7 This double use of notation is unproblematic as it is always clear from the context whether we mean
the experimenter or the corresponding experiment.

13

corresponds, up to relabelings, to a Repeated Measurement Experiment, as introduced

in Section 3.1, and can therefore be described using standard quantum theory. Note that
such a description is not possible for the joint experiment, since z and w, for instance,
cannot simultaneously be regarded as observables according to quantum theory.
Definition of Experiment F1
The experimenter F1 prepares the state S of S depending on the value r. Furthermore,
we assume that she is interested in the outcome w measured at time t = n:40. The
corresponding event space may therefore be defined as

[F1] = (t, r, S , w) T {head, tail, } HS {ok, fail, } :

if t = n:10 then r = head S = |i and r = tail S = |i ,
(11)
for T as defined by (3). To formalise that w is the outcome of a quantum measurement, we relate this experiment to one of the set RMEHS ,{wH } , where HS is the Hilbert
H
are measurement projectors in the
space associated to the electron spin S and where w
n:20
Heisenberg picture. Specifically, taking U = U
(for any n N0 ) to be an isometry
from S to F2 such that
U |iS = | 12 iF2

and

U |iS = |+ 12 iF2 ,

H
= U |wihw|U for w {ok, fail}. We then require that
we define w

E RMEHS ,{wH } :

(t, , S , w) sF1 (t0 , S , w) sE ,

(12)

where the mapping t 7 t0 is such that n:10 7 n:00, n:40 7 n:01, and n:50 7 n:02.
Note that this mapping must not be injective. While the time information contained in
the plot sE is thus more coarse-grained than that of sF1 , it is still sufficient to apply the
laws of quantum theory, as formulated by (QT).
Definition of Experiment F2
Experimenter F2 measures the vertical spin direction z. We assume that she is also
interested in the state S as well as the randomness r it depends on. We therefore define
the event space as

[F2] = (t, r, S , z) T {head, tail, } HS { 21 , + 12 , } :

if t = n:10 then r = head S = |i and r = tail S = |i .
(13)
To specify how z arises as the outcome of a quantum measurement, we define zH = |zihz|
for z { 12 , + 12 } and demand that
E RMEHS ,{zH } :

(t, , S , z) sF2 (t0 , S , z) sE ,

where t 7 t0 is such that n:10 7 n:00, n:20 7 n:01 and n:40 7 n:02.

14

(14)

Definition of Experiment A
Experimenter A is measuring x and, in addition, interested in the outcome z of the
measurement at t = n:20. Since these outcomes ultimately depend on the initial state
C HC of the quantum coin C, we also include it in the event space, which we choose
to be

[A] = (t, C , z, x) T HC { 21 , + 21 , } {ok, fail, } :
if t = n:00 then C = C0

(15)

for C0 as defined by (10). To describe this as a quantum measurement, let V = V n:00n:10

be an isometry from C to F1 S such that
V |headiC = |headiF1 |iS

and

V |tailiC = |tailiF1 |iS

H
and define the projectors z,x
= V |xihx| |zihz|V . We then require that
H } :
E RMEHC ,{z,x

(16)

if t = [n:00, n:20) :

(t, C , , ) sA

if t = [n:20, n:30) :

(t, C , z, ) sA

if t = [n:30, n:40) :

(t, C , z, x) sA

if t = [n:40, n:50) :

(t, C , , x) sA

n:00, C , sE


n:01, C , (z, ) sE


n:01, C , (z, x) sE


n:01, C , (, x) sE ,

where we restrict to values z 6= and x 6= .

Definition of Experiment W
Experimenter W measures w and is also interested in x. Similarly to the above, the event
space shall therefore be

[W] = (t, C , x, w) T HC {ok, fail, } {ok, fail, } :

if t = n:00 then C = C0 .
(17)
To describe this as a quantum measurement, we define the projectors


H
x,w
= V |xihx| U |wihw|U V ,
where U and V are the isometries from above. We then require that
H } Obs :
E RMEHC ,{x,w

(18)

if t = [n:00, n:30) :

(t, C , , ) sW

n:00, C , sE


n:01, C , (x, ) sE


n:01, C , (x, w) sE ,

if t = [n:30, n:40) :

(t, C , x, ) sW

if t = [n:40, n:50) :

(t, C , x, w) sW

where we restrict to values x 6= and w 6= . Note that E Obs ensures that the
single-world property (SW) holds from the viewpoint of W. The condition that the
15

experiment is repeated until the halting criterion is satisfied (which can be tested by W)
then corresponds to (9), with z = (ok, ok). This could also be restated as
n = 0 or ((n 1):40, , x, w) sW for (x, w) 6= (ok, ok)
=

t [n:00, n:40] : (t, , , ) sW . (19)

By induction, it is easy to see that this condition implies that

t [n:00, n:40] : (t, , , ) sW

(20)

for any n not larger than the minimum one satisfying (n:40, , x = ok, w = ok) sW .
Compatibility conditions
The experiments F1, F2, A, and W obviously have certain overlapping elements after
all, they are all part of one big experiment. As explained in Section 2.3, the overlap
between the experiments can be defined via compatibility constraints. Specifically, we
demand that
(t, , , w) sF1

(t, , , w) sW ,

(21)

which models that the quantity w defined in the experiment W is the same as the one F1
refers to. Similarly, the compatibility constraints which model that r, z, and x denote
the same quantities in each experiment can be written as
(t, r, , ) sF1

(t, r, , ) sF2
A

(22)

F2

(t, , , z) s

(t, , z, ) s

(23)

(t, C , , x) sA

(t, C , x, ) sW .

(24)

Proof

We split the proof of Theorem 1 in two parts. In the first, we use property (QT) to
analyse the sub-experiments F1, F2, A, and W separately. We recall that each of them
models a part of the Extended Wigners Friend Experiment that is of interest to one of
the four experimenters. Accordingly, the conclusions we obtain from this analysis, (25),
(26), (29), and (31), are those that the experimenters would arrive at if they applied
standard quantum theory. Then, in the second part of the proof, we use properties (SW)
and (SC) to combine these individual conclusions, which then leads to a contradiction.

5.1

Analysis of individual views

For the following, let T be any theory that satisfies property (QT) and let s be any story
that is not forbidden by T .

16

Analysis of Experiment F1
By linearity, we have
U |iS = U

1/2|i
S

1/2|i
S

= |failiF2 ,

H
which immediately implies that kfail
|iS k = 1. Hence, item (a) of property (QT), which
applies due to (12), tells us that

(n:10, , |iS , ) sF1

(n:40, , , w = fail) .

Using furthermore the constraint (11), we conclude that

(n:10, r = tail, , ) sF1

(n:40, , , w = fail) sF1 .

(25)

Analysis of Experiment F2
Here (QT) applies because of (14). Item (b) thus asserts that
(n:20, , , z = + 21 ) sF2

H
S : (n:10, , S , ) sF2 and k+
6 0.
1 S k =

By the constraint (13), we conclude that

(n:20, , , z = + 12 ) sF2

(n:10, r = tail, , ) sF2 .

(26)

Analysis of Experiment A
Using the explicit form (10) of C0 we find that
p
p
V C0 = 1/3|headiF1 |iS + 2/3|tailiF1 |iS
p
p
p
= 1/3|headiF1 |iS + 1/3|tailiF1 |iS + 1/3|tailiF1 |iS
p
p
= 2/3|failiF1 |iS + 1/3|tailiF1 |iS .

(27)

This vector is obviously orthogonal to |okiF1 |iS . This means that C0 is orthogonal
H
to
. Furthermore, (15) guarantees that
1
,ok
2

(n:00, C , , ) sA

C = C0 .

Let E be an experiment as defined by (16). It follows from item (b) of (QT) that


n:01, , (z = 12 , x = ok)
/ sE .

(28)

By (16) we also have

(n:40, , , x = ok) sA

(n:01, , (z, x = ok)) sE

=

(n:20, , z, ) sA

for some value z 6= 0. Since, according to (28), z 6= 21 , we conclude that

(n:40, , , x = ok) sA

=
17

(n:20, , z = + 12 , ) sA .

(29)

Analysis of Experiment W
Using (27) we find
U V C0 =

2/3

|failiF1 | 12 iF2 +

1/3

|tailiF1 |+ 21 iF2 .

(30)

The overlap between this state and the state |okiF1 |okiF2 equals
p
p
1/3hok|taili hok|+ 1 i
1/12 .
F1
2 F2 =
H
In other words, the state C0 has a non-zero overlap with the projector ok,ok
. Furthermore, (17) ensures that the system is always prepared in this state, unless no state was
prepared at all. Item (c) of (QT), which applies due to (18), together with the requirement (19) that the experiment is repeated until the halting condition is satisfied, implies
that there must exist a round n where x = w = ok occur, i.e.,

(n:40, , x = ok, w = ok) sW

(31)

holds whenever the plot sW is defined.

5.2

Combining the views

Assume by contradiction that T satisfies the three properties (QT), (SW), and (SC).
Property (SC) implies that there must exist a story s that is not forbidden by T such
that all of sF1 , sF2 , sA , and sW are defined. In the following we consider such a story s.
According to the analysis above, there must exist a round n such that (31) holds.
Taking n to be the smallest such number, it follows from (20) that
t [n:00, n:40] : (t, , , ) sW .
Furthermore, by virtue of (18) we can apply property (SW), which implies


 (x, w) {ok, fail} {ok, fail} : (n:40, , x, w) sW  = 1 .

(32)

(33)

Since we have chosen n such that (31) holds, it follows that

(n:40, , x, w) sW

x = w = ok .

(34)

Using the compatibility condition (24) we obtain

(n:40, , , x = ok) sA .
Constraint (29), which resulted from the quantum-mechanical analysis from As viewpoint, hence implies that
(n:20, , z = + 21 , ) sA .
We now use the compatibility condition (23) to infer that
(n:20, , z = + 21 , ) sF2 .
18

Applying (26), which resulted from the quantum-mechanical analysis from F2s viewpoint, we obtain
(n:10, r = tail, , ) sF2 .
Using the compatibility criterion (22), this means that
(n:10, r = tail, , ) sF1 .

(35)

The quantum-mechanical analysis from F1s viewpoint, (25), then gives

(n:40, , , w = fail) sF1 .
By the compatibility condition (21) this means that
(n:40, , , w = fail) sW .
But this is in contradiction to (34), which concludes the proof of Theorem 1.

Possible scenarios

According to Theorem 1, if one wishes to devise a physical theory, one has to give up at
least one of the assumptions (QT), (SW), or (SC). In the following three subsections, we
discuss possible scenarios that arise when one of the assumptions is abandoned. They
usually correspond to theories that have been proposed in the literature. The fourth
subsection addresses the question whether there are other implicit assumptions built
into the framework used here.

6.1

Theories that violate (QT)

Property (QT) demands that the laws of quantum theory are valid even for systems
that are complex enough to contain an observer (who herself applies quantum theory
to describe a system in her possession). This assumption is generally violated by theories that postulate a modification of the usual Schrodinger equation [GR95, Wei12].
Examples include spontaneous collapse models such as the GRW theory and extensions
thereof [GRW86, Pea89, Tum06], as well as gravity induced collapse models [Kar66,
Di
o89, Pen96] (see [BLS+ 13] for a review). The same is true for the proposal of [AG15]
to supplement the basic formalism of quantum theory with the postulate that measurements must be carried out in a context (which may include the measurement device).
Since, according to the postulate, this context cannot be treated itself as a quantum
system, it rules out the possibility of applying quantum measurements to measurement
devices, thus violating (QT).
Conversely, property (QT) is not usually violated by hidden variable theories. The
reason is that (QT) only refers to statements that can be made with certainty within
quantum theory. In particular, it does not impose any constraints on the statistics of
measurement outcomes, and hence it does not rule out theories that provide more informative predictions than quantum theory [CR11]. To illustrate this, one may take
a variant of Bohmian mechanics where the particle positions are known. This theory,
although being deterministic, would never predict an outcome that is forbidden by quantum theory, and hence still satisfy (QT).
19

+|

Figure 4: The intuition behind many-worlds. Slightly varying the Schrodingers cat
gedankenexperiment [Sch35], a mechanism may decide, based on the outcome of a spin
measurement, whether or not the cat is fed. According to a many-worlds interpretation,
the two perceptions of the cat, hungry and happy, are equally real. While this may
sound counter-intuitive, a much more familiar situation is obtained when the feeding
mechanism is replaced by one triggered by time. Assuming the cat has no good memory,
the situation remains the same for her: the two perceptions, hungry and happy, are
equally real.

6.2

Theories that violate (SW)

Property (SW) demands that only one single outcome occurs when we measure a quantum system. It captures the intuition that the outcome we are aware of is the only
real one. Probably the most prominent representative of a theory which abandons this
intuition is the relative state formalism, which is also known as the many-worlds or the
Everett interpretation [Eve57, Whe57, DeW70, Deu85, Deu97, Vai16]. It proclaims that
the measurement of a quantum system results in a branching into different worlds,
in each of which one of the possible measurement outcomes occurs. The different outcomes are therefore all equally real, corresponding to statement (S2) in the case of our
introductory example of a spin measurement (cf. Fig. 4).
Various variants and extensions of the relative state formalism have been proposed.
Among them are the many-minds interpretation [Zeh70, AL88] and the parallel lives
theory [BRR13], as well as notions such as quantum Darwinism [Zur07]. Their common
feature is that they do not postulate a physical mechanism that singles out one particular
measurement outcome, although observers have the perception of single outcomes. They
are therefore all examples of theories that are not single-world in the sense of (SW).
We recall, however, that property (SW) does not demand that one can simultaneously
assign unique outcomes to any measurement made during an experiment. Rather, the
requirement is that the measurement outcomes accessible to one observer (in our case
20

experimenter W) have single values. Our analysis of the Extended Wigners Friend
Experiment therefore leads to a conclusion that differs, for instance, from that of [Bru15],
who argued that facts cannot exist per se, but that they may still exist relative to
observers. Theorem 1 also excludes this latter possibility.

6.3

Theories that violate (SC)

It is a consequence of Theorem 1 that deterministic hidden variable theories must violate

property (SC), i.e., they cannot be self-consistent. Indeed, as argued in Section 6.1, they
satisfy (QT), and because they are (by definition) deterministic, measurements have only
one single outcome, so that (SW) holds.
An example of such a hidden variable theory is Bohmian mechanics. The conclusion that Bohmian mechanics is not self-consistent may be compared to the results
of [CM02, KW10]. They suggest that, in an experiment with entangled particles, the
behaviour of the Bohmian particle positions is not consistent with the statistics that
quantum theory would predict for position measurements. However, since Bohmian
particle positions cannot in general be identified with the outcomes of position measurements [ESSW92, Vai05], this finding does not point to a fundamental inconsistency of
Bohmian mechanics [Gis15]. This is in contrast to the conclusions we can draw from
Theorem 1, which shows that an inconsistency arises independently of how one interprets
the Bohmian particle positions.
It is certainly unsatisfactory if a theory is not self-consistent. One may therefore ask
whether there is an easy fix. One possibility could be to restrict the range of applicability
of the theory and add the rule that its predictions are only valid if an experimenter who
makes the predictions keeps all relevant information stored. While we do not normally
impose such a rule when using theories to make predictions, this would, at least in the
case of Bohmian mechanics, remove the inconsistency. Indeed, since in the Extended
Wigners Friend Experiment F1 may not be able to store the value r until t = n:40, the
theory could no longer be applied to analyse her part of the experiment. Formally, this
means that (12) would no longer be a valid requirement and, hence, (25) could not be
established.

6.4

Relaxing the notion of a physical theory

The formulation of Theorem 1 is based on a framework, as outlined in Section 2, which

allows us to reason about physical theories. Hence, any structure imposed by the framework potentially limits the generality of the claim. One may therefore try to further relax
it. However, compared to other frameworks used in the literature on the foundations of
quantum theory, the one used here is already rather minimalistic. In particular, it is not
demand that a physical theory provides predictions that can be expressed in terms of
probability distributions.
To illustrate this, it is instructive to have a quick look at other approaches, such as
the one taken by Bell [Bel66]. The assumptions that enter Bells theorem are usually
formulated in terms of probability distributions of certain random variables that model
the outcomes of measurements. The model therefore comes with the a priori assumption
that all measurements, even if they are carried out by different experimenters, have a welldefined outcome (the value of the random variable), and that a physical theory allows us

21

to calculate predictions (the probability distributions assigned to these variables). The

same holds true for various results based on probabilistic frameworks [Bar07, BW16] (cf.
[Har11, CDP11, MM11, CR12, PBR12, OCB12, LS13] for some recent examples). In
contrast to this, Theorem 1 completely avoids the use of the notion of probability, so
that no such assumption is necessary.
One may now ask whether there are still other assumptions built into the framework
used here. For example, the idea of considering multiple experimenters, which is central
to our argument, could be problematic if one takes a more radical subjective viewpoint.
A nice example is QBism, according to which probabilities represent beliefs of an agent
and are therefore entirely subjective [FMS14]. Even if an agent assigns probability 1 to
a particular outcome of a measurement, this should not be taken as a fact, and QBism
would even allow that another agent assigns probability 0 to the same outcome. This
suggests that, within QBism, it may not be sensible to carry out an analysis that is
based on the combination of different experimenters views. Nevertheless, since it is
rather common that we try to infer the behaviour of others by reasoning about their
decision-making, it could still be interesting to explore how QBism should be applied
to situations where one agent uses QBism to express his believes about another agents
actions who herself applies QBism. We leave a study of the Extended Wigners Friend
Experiment from such a Bayesian viewpoint to future research.

Conclusions

A natural requirement to any reasonable physical theory T is that different observers

who apply T should not arrive at logically contradicting statements. This is captured
by property (SC), which demands that there exists at least one possible story s that
none of the different observers would consider as forbidden. Once one accepts (SC) as
an unavoidable requirement, our main result, Theorem 1 implies that we are left with
the option to either deny (QT) or (SW).
It is in principle possible to decide between these two options by an appropriately
designed experiment. The experiment that would need to be carried out is a test whether
the predictions of standard quantum theory are valid for systems that are complex enough
to count as observers. As famously noted by Bell, defining precisely what this means
appears to be impossible [Bel90]. However, experiments that test quantum coherence
for macroscopic systems [AH14] certainly provide evidence in favour of (QT), and hence
against (SW).
While the ultimate experiment would involve human observers, one could argue that
it is sufficient to consider systems that have the same level of complexity. This could
be achieved once we are able to build scalable quantum computers.8 Theorem 1 therefore provides another incentive to build such machines. If they work according to the
predictions of quantum theory, we know we are left with Scenario 2 described in the
introduction. That is, we are forced to give up the view that there is one single reality.
8 Scalability means that arbitrarily complex computing devices can be built from elementary components by composition.

22

Acknowledgments
We would like to thank Alexia Auff`eves, Serguei Beloussov, Hans Briegel, Ldia del Rio,
David Deutsch, Artur Ekert, Nicolas Gisin, Philippe Grangier, Thomas M
uller, Sandu
Popescu, R
udiger Schack, and Vlatko Vedral for discussing ideas that led to this work.
This project was supported by the Stellenbosch Institute for Advanced Study (STIAS),
by the Swiss National Science Foundation (SNSF) via the National Centre of Competence
in Research QSIT, by the European Research Council (ERC) via grant No. 258932,
and by the European Commission via the project RAQUEL.

References
[AG15]

A. Auff`eves and P. Grangier. Contexts, systems and modalities: A new ontology for
quantum mechanics. Foundations of Physics, 46:121137, 2015.

[AH14]

M. Arndt and K. Hornberger. Testing the limits of quantum mechanical superpositions. Nat. Phys., 10:271277, 2014.

[AL88]

D. Albert and B. Loewer. Interpreting the many worlds interpretation. Synthese,

77:195213, 1988.

[Bar07]

J. Barrett. Information processing in generalized probabilistic theories. Phys. Rev.

A, 75:032304, 2007.

[Bel66]

J.S. Bell. On the problem of hidden variables in quantum mechanics. Rev. Mod.
Phys., 38:447452, 1966.

[Bel90]

J.S. Bell. Against measurement. Phys. World, 3:33, 1990.

[BLS 13] A. Bassi, K. Lochan, S. Satin, T.P. Singh, and H. Ulbricht. Models of wave-function
collapse, underlying theories, and experimental tests. Rev. Mod. Phys., 85:471527,
2013.
[Boh52]

D. Bohm. A suggested interpretation of the quantum theory in terms of hidden

variables. I. Phys. Rev., 85:166179, 1952.

[BRR13]

G. Brassard and P. Raymond-Robichaud. Is Science Compatible with Free Will?

Exploring Free Will and Consciousness in the Light of Quantum Physics and Neuroscience, chapter Can Free Will Emerge from Determinism in Quantum Theory?
Springer, 2013.
Brukner. On the quantum measurement problem. arXiv:1507.05255, 2015.
C.

[Bru15]
[BW16]

H. Barnum and A. Wilce. Quantum Theory: Informational Foundations and Foils,

chapter Post-Classical Probability Theory, pages 367420. Springer, Dordrecht, 2016.

[CDP11]

G. Chiribella, G.M. DAriano, and P. Perinotti. Informational derivation of quantum

theory. Phys. Rev. A, 84:012311, 2011.

[CM02]

M. Correggi and G. Morchio. Quantum mechanics and stochastic mechanics for

compatible observables at different times. Ann. Phys., 296:371389, 2002.

[CR11]

R. Colbeck and R. Renner. No extension of quantum theory can have improved

predictive power. Nat. Commun., 2:411, 2011.

[CR12]

R. Colbeck and R. Renner. Is a systems wave function in one-to-one correspondence

with its elements of reality? Phys. Rev. Lett., 108:150402, 2012.

[DB27]

L. De Broglie. La mecanique ondulatoire et la structure atomique de la mati`ere et

du rayonnement. J. Phys. Radium, 8:225241, 1927.

23

[Deu85]

D. Deutsch. Quantum theory as a universal physical theory. Int. J. Theor. Phys.,

24:141, 1985.

[Deu97]

D. Deutsch. The Fabric of Reality: The Science of Parallel Universes and Its Implications. Allen Lane Science, 1997.

[DeW70]

B.S. DeWitt. Quantum mechanics and reality. Phys. Today, 23:155165, 1970.

[Di
o89]

L. Di
osi. Models for universal reduction of macroscopic quantum fluctuations. Phys.
Rev. A, 40:11651174, 1989.

[DT09]

D. D
urr and S. Teufel. Bohmian Mechanics: The Physics and Mathematics of Quantum Theory. Springer, 2009.

[ESSW92] B.-G. Englert, M.O. Scully, G. S

ussmann, and H. Walther. Surrealistic Bohm trajectories. Z. Naturforsch., 47:11751186, 1992.
[Eve57]

H. Everett. Relative state formulation of quantum mechanics. Rev. Mod. Phys.,

29:454462, 1957.

[FMS14]

C.A. Fuchs, N.D. Mermin, and R. Schack. An introduction to QBism with an

application to the locality of quantum mechanics. Am. J. Phys., 82:749754, 2014.

[Gis15]

N. Gisin. Why Bohmian mechanics? One and two-time position measurements, Bell
inequalities, philosophy and physics. arXiv:1509.00767, 2015.

[GR95]

N. Gisin and M. Rigo. Relevant and irrelevant nonlinear Schr

odinger equations. J.
Phys. A, 28:7375, 1995.

[GRW86] G. C. Ghirardi, A. Rimini, and T. Weber. Unified dynamics for microscopic and
macroscopic systems. Phys. Rev. D, 34:470491, 1986.
[Har92]

L. Hardy. Quantum mechanics, local realistic theories, and Lorentz-invariant realistic

theories. Phys. Rev. Lett., 68:29812984, 1992.

[Har93]

L. Hardy. Nonlocality for two particles without inequalities for almost all entangled
states. Phys. Rev. Lett., 71:16651668, 1993.

[Har11]

L. Hardy. Reformulating and reconstructing quantum theory. arXiv:1104.2066, 2011.

[Kar66]

F. Karolyhazy. Gravitation and quantum mechanics of macroscopic objects. Il Nuovo

Cimento A (1971-1996), 42:390402, 1966.

[KW10]

J. Kiukas and R.F. Werner. Maximal violation of Bell inequalities by position measurements. J. Math. Phys., 51, 2010.

[LS13]

M.S. Leifer and R.W. Spekkens. Towards a formulation of quantum theory as a

causally neutral theory of Bayesian inference. Phys. Rev. A, 88:052130, 2013.

[MM11]

L. Masanes and M.P. M

uller. A derivation of quantum theory from physical requirements. New J. Phys., 13:063001, 2011.
Brukner. Quantum correlations with no causal order.
O. Oreshkov, F. Costa, and C.
Nat. Commun., 3:1092, 10 2012.

[OCB12]
[PBR12]

M.F. Pusey, J. Barrett, and T. Rudolph. On the reality of the quantum state. Nat.
Phys., 8:475478, 2012.

[Pea89]

P. Pearle. Combining stochastic dynamical state-vector reduction with spontaneous

localization. Phys. Rev. A, 39:22772289, 1989.

[Pen96]

R. Penrose. On gravitys role in quantum state reduction. Gen. Rel. Gravit., 28:581
600, 1996.

[Sch35]

E. Schr
odinger. Die gegenw
artige Situation in der Quantenmechanik. Naturwissenschaften, 23:823828, 1935.

24

[Tum06]

R. Tumulka. On spontaneous wave function collapse and quantum field theory. In

Proc. R. Soc. A, volume 462, pages 18971908. The Royal Society, 2006.

[Vai05]

L. Vaidman. The reality in Bohmian quantum mechanics or can you kill with an
empty wave bullet? Found. Phys., 35:299312, 2005.

[Vai16]

L. Vaidman. Many-worlds interpretation of quantum mechanics. In The Stanford

Encyclopedia of Philosophy. Stanford University, 2016.

[Wei12]

S. Weinberg. Collapse of the state vector. Phys. Rev. A, 85:062116, 2012.

[Whe57]

J.A Wheeler. Assessment of Everetts relative state formulation of quantum theory.

Rev. Mod. Phys., 29:463, 1957.

[Wig67]

E.P. Wigner. Symmetries and Reflections, chapter Remarks on the Mind-Body Question, pages 171184. Indiana University Press, 1967.

[Zeh70]

H.D. Zeh. On the interpretation of measurement in quantum theory. Found. Phys.,

1:6976, 1970.

[Zur07]

W.H. Zurek. Relative states and the environment: einselection, envariance, quantum
Darwinism, and the existential interpretation. arXiv:0707.2832, 2007.

25